Dissimilatory Nitrate Reduction to Ammonium and Responsible Microbes in Japanese Rice Paddy Soil

Nitrification–denitrification processes in the nitrogen cycle have been extensively examined in rice paddy soils. Nitrate is generally depleted in the reduced soil layer below the thin oxidized layer at the surface, and this may be attributed to high denitrification activity. In the present study, we investigated dissimilatory nitrate reduction to ammonium (DNRA), which competes with denitrification for nitrate, in order to challenge the conventional view of nitrogen cycling in paddy soils. We performed paddy soil microcosm experiments using 15N tracer analyses to assess DNRA and denitrification rates and conducted clone library analyses of transcripts of nitrite reductase genes (nrfA, nirS, and nirK) in order to identify the microbial populations carrying out these processes. The results obtained showed that DNRA occurred to a similar extent to denitrification and appeared to be enhanced by a nitrate limitation relative to organic carbon. We also demonstrated that different microbial taxa were responsible for these distinct processes. Based on these results and previous field observations, nitrate produced by nitrification within the surface oxidized layer may be reduced not only to gaseous N2 via denitrification, but also to NH4+ via DNRA, within the reduced layer. The present results also indicate that DNRA reduces N loss through denitrification and nitrate leaching and provides ammonium to rice roots in rice paddy fields.

Rice is one of the most important agronomic products in the world. The supply of nitrogen (N) to rice roots generally limits production. Various biochemical processes involved in N cycling occur in paddy soils. As initially discovered in the 1930s, nitrification (NH 4 + →NO 2 -→NO 3 -) occurs within a thin oxidized soil surface layer, while denitrification (NO 3 →NO 2 -→NO→N 2 O→N 2 ) occurs within a reduced soil layer below the oxidized layer . Nitrification-denitrification processes form the core of the N cycle, and, thus, have been extensively examined (see reviews by Ishii et al., 2011). The diverse microbes responsible for these processes were identified in the 2000s and 2010s Yoshida et al., 2009;Ishii et al., 2010;Yoshida et al., 2010;Tago et al., 2011;Yoshida et al., 2012;Nishizawa et al., 2013;Wei et al., 2015;Masuda et al., 2017;Fujimura et al., 2020). Since rice preferably utilizes NH 4 + over NO 3 - (Sasakawa and Yamamoto, 1978), the sequential processes of nitrification-denitrification that eventually lead to N loss from soil appear to reduce the efficiency of ammonium-based fertilization.
Submerged anaerobic paddy soils are generally NO 3 -depleted environments (Nishimura et al., 2004;Itoh et al., 2013), which is in contrast to the continuous presence of dissolved organic carbon (C) in these soils (Xu et al., 2013;Said-Pullicino et al., 2016). This may be due to the strong denitrification activity of these soils ; however, DNRA may be an alternative process that depletes NO 3 -in paddy soils. Denitrification functions as an N loss process. In contrast, DNRA functions as a N retention process by preventing N loss via denitrification and reducing water-leachable NO 3 -to NH 4 + , which is more likely to be absorbed by negatively charged soil particles (Rütting et al., Article ME20069 2011). DNRA also supplies NH 4 + to rice roots in paddy soils. DNRA has been detected in various soil environments, including forests (Zhang et al., 2018), grasslands (Rütting and Müller, 2008), upland croplands (Putz et al., 2018), and deserts (Yang et al., 2017). However, research on DNRA in rice paddy fields remains limited, in terms of the number of studies conducted and the range of geographic locations examined (USA, China, Australia, and Malaysia) (Buresh and Patrick, 1978;Shan et al., 2016;Pandey et al., 2018;. Furthermore, microbial populations actively carrying out DNRA in paddy fields have not been characterized in detail. In the present study, we initially investigated whether DNRA occurs in Japanese rice paddy soil and then examined the effects of a NO 3 -limitation relative to organic car--bon on the partitioning of NO 3 reduction between DNRA and denitrification. We then identified the microbial populations carrying out DNRA and denitrification to establish whether these processes were driven by similar or different microbial communities.

Materials and Methods
We collected two soil samples for soil microcosm experiments. Both samples were collected from a paddy field at the Niigata Agricultural Research Institute (37°43'N, 138°87'E) in Niigata Prefecture, the richest rice-producing area in Japan. Rice (Oryza sativa L., cv. Koshihikari) has been cultivated in the field since 2003. The soil is classified as grey lowland soil. We sampled soil at a depth of 0-20 cm, removed the oxidized layer (1 cm from the top), then sieved (<4 mm) and mixed the soil. The biogeochemical conditions of the field were previously described by Itoh et al. (2013). The sample collected on June 11, 2014 was reserved for the first experiment, while that collected on June 12, 2015 was used in the second and third experiments (see below). The concentrations of dissolved organic carbon and NH 4 + in the sample collected on June 12, 2015 were 25.9±0.3 μg C (g soil) -1 (n=5), as measured by a TOC meter (TOC-VCPH, Shimadzu), and 1.4±0.1 μg N (g soil) -1 (n=5), as measured by the indophenol spectrophotometric method with a 2 M KCl extract (Isobe et al., -2011b), whereas NO 3 was not detected. These data were not obtained for the sample collected on June 11, 2014.
Our first objective was to investigate the presence of DNRA in Japanese rice paddy soil. Ten grams of soil and 9.5 mL of water were dispensed into a 50-mL glass vial. The vial was crimp-sealed using a butyl stopper and the headspace was replaced by pure helium three times.  (Tiedje, 1994). We then added 32 mL of 3 M KCl to the vial, shook at 200 rpm for 1 h, and filtered the extract using a glass filter --(GF/F). We measured 15 NO 3 , 15 NO 2 , and 15 NH 4 + in the extract using the method described by Isobe et al. (2011b) We also prepared soil microcosms for the gene transcription survey. RNA was extracted from a 2-g soil sample and purified, and cDNA was synthesized with RNA solution as previously described (Wei et al., 2015). We used soil samples before the addition of 14 NH 4 15 NO 3 and incubated the vials for 8 and 12 h. The gene transcripts of pentaheme nitrite reductase (NrfA), which catalyzes the last step of DNRA (NO 2 -→NH 4 + ), were amplified using a polymerase chain reaction (PCR) with the primers nrfAF2aw and nrfAR1 (Welsh et al., 2014). These primers and PCR conditions are listed in Table  S1. The composition of the reaction mixture and the details of the procedure are described in Isobe et al. (2012).
-Our second objective was to examine the effects of the NO 3 limitation on DNRA. We followed the same procedure as the first experiment, except for the org-C/NO 3 -of 14 NH 4 15 NO 3 , glucose sol--utions, and incubation times. Org-C/NO 3 (mol/mol) ranged from 0, 1, 3, 5, and 7 to 9 by changing the concentration of added glucose (from 0, 1. Our final objective was to identify the microbial populations involved in DNRA and denitrification. We constructed clone libraries of the gene transcripts for DNRA and denitrification. RNA was extracted from the soils of org-C/NO 3 -=3 (see below) used in the second experiment. The nrfA gene transcripts were amplified by PCR using the primers nrfAF2aw and nrfAR1. The gene transcripts of copper-containing nitrite reductase (NirK) and cytochrome cd1-containing nitrite reductase (NirS), which catalyze the step of denitrification (NO 2 -->NO), were also amplified using the following primers: nirKC1F and nirKC1R (nirK Cluster I), nirKC2F and nirKC2R (nirK Cluster II), and nirSC1F and nirSC1R (nirS Cluster I). Denitrifiers may be phylogenetically grouped based on their nirK and nirS sequences into several clusters. We previously demonstrated that denitrifiers with nirK Cluster I and Cluster II and nirS Cluster I were dominant in the soils tested (Wei et al., 2015;Fujimura et al. 2020), and, thus, targeted these genes. The primers and PCR conditions used are described in Table S1. The composition of the reaction mixture and the details of the procedure were previously described by Wei et al. (2015). Cloning of the PCR amplicons and sequencing were performed as described previously (Isobe et al., 2012). After removing the primer-annealing regions, nucleotide sequences were translated into amino acid sequences using EMBOSS (Rice et al., 2000). We developed a reference database of nrfA, nirK, or nirS amino acid sequences by curating the nrfA, nirK, and nirS databases downloaded on February 20, 2020 from the FunGene database (Fish et al., 2013). We used the BLASTP function of the BLAST+ tool (Camacho et al., 2009) for the homology search using the following threshold values: E value <0.001 and percent identity >60% for the taxonomic assignment. We then constructed the phylogenetic trees of clone sequences with the reference sequences. These sequences were aligned using Clustal W version 2.0, and maximum likelihood trees were constructed using a bootstrap analysis (100 replicates) performed in MEGA X (Kumar et al., 2018). We also investigated the presence of the nrfA, nirK, and nirS genes within the complete genome of the assigned genus using the BLASTP search based on the genome sequences deposited in the Kyoto Encyclopedia of Genes and Genomes (KEGG) GENOME database (Kanehisa et al., 2017).

Nucleotide sequence accession numbers
The nucleotide sequence data reported are available in the DDBJ Sequenced Read Archive under the accession numbers LC532390-LC532446 for nrfA, LC532447-LC532534 for nirK, and LC532535-LC532580 for nirS.

DNRA activity in paddy soil
The results obtained from our first experiment confirmed that DNRA occurred in rice paddy soils at a similar rate to denitrification (Fig. 1A). NO 3 was eventually reduced to NH 4 + and N 2 in soils amended with glucose. The reduction to NO 2 was also observed after 4-12 h, whereas the concentration of N 2 O remained low around the detection limit. NO 3 -reduction proceeded even without the addition of glucose and mostly led to N 2 production. The extent of NH 4 + production via DNRA with glucose during 12 h was similar to that of N 2 production via denitrification, whereas NH 4 + production via DNRA was minimal without glucose. We did --not detect the oxidation of NH 4 + to NO 2 or NO 3 in soil microcosms amended with 15 NH 4 NO 3 and glucose or with 15 NH 4 NO 3 alone. nrfA was strongly expressed in soils amended with glucose after 8 h when 15 NH 4 + was produced at the highest rate (Fig. 1B). Conversely, nrfA expression was less evident before NH 4 + production started or when it ended (after 12 h). nrfA expression was also less evident in soils without glucose. These results indicate that DNRA occurred in rice paddy soils and was driven by nrfApossessing microbial populations.

Effects of the nitrate limitation relative to organic carbon on DNRA and denitrification
The results from our second experiment showed that the ratio of organic carbon to NO 3 (org-C/NO 3 -) affected the partitioning of NO 3 -reduction processes; higher org-C/NO 3 -enhanced DNRA over denitrification ( Fig. 2A and B). NO 3 reduction and NH 4 + and N 2 production progressed as org---C/NO 3 increased. NO 2 and N 2 O concentrations were the highest at org-C/NO 3 -of 3. The ratio of DNRA to denitrifi--cation also became higher as org-C/NO 3 increased from -10% with C-glucose/NO 3 of zero, to 50% with C--glucose/NO 3 of 9. We did not detect NH 4 + oxidation to NO 2 -or NO 3 -in soil microcosms amended with 15 NH 4 NO 3 and glucose or with 15 NH 4 NO 3 alone These results indicate . that DNRA is more likely to occur under NO 3 -limiting (relative to org-C) environments. The ratios of DNRA to denitrification differed between the first and second experiments: the ratios of soil samples with org-C/NO 3 -of 0 and 9 were larger and smaller in the second experiment, respectively. Although the reason for this result currently remains unclear, it may depend on the sampling period (see above).

Microbial populations carrying out DNRA and denitrification
Since DNRA and denitrification both occurred simultane--ously and actively in soils with org-C/NO 3 of 3, we analyzed the microbial populations responsible for these processes based on the sequences of nrfA transcripts for DNRA and those of nirK and nirS transcripts for denitrification. We obtained 57, 43, 45, and 46 clones for nrfA, nirK in Cluster I, nirK in Cluster II, and nirS in Cluster I, respectively. The taxonomic composition of DNRA and denitrifying microbes appeared to differ (Table 1 and S2). Many (43.9%) of the nrfA sequences showed the highest similarity to those from the genera Anaeromyxobacter and Geobacter (class Deltaproteobacteria), while many others showed the highest similarity to those from the taxa of which genuslevel taxonomy is unknown. In contrast, nirK sequences in Cluster I showed the highest similarity to those from the genera Bradyrhizobium and Afipia in the Bradyrhizobiaceae family (class Alphaproteobacteria). nirK sequences in Cluster II showed the highest similarity to those from the genera Rhodanobacter (class Gammaproteobacteria) and Ralstonia (class Betaproteobacteria). nirS sequences showed the highest similarity with the genera Thiothrix, Rhodanobacter, and Marinobacter (class Gammaproteobacteria) and the genus Burkholderiaceae (class Betaproteobacteria). Although the majority of nrfA, nirS, and nirK sequences were assigned to the reference sequences with high similarity, they did not form distinct phylogenetic clades with these reference sequences with a bootstrap support of >80%, except for the clones assigned to the genus Geobacter or Pseudogulbenkiania (Fig. S1), which may have been due to the lack of nrfA, nirS, and nirK reference sequences within the database preventing accurate taxonomic assignments. Conversely, complete genome sequences within the assigned genus with nrfA, such as Geobacter and Anaeromyxobacter, did not possess nirS or nirK, whereas those with nirK or nirS did not possess nrfA (Table S3). These results suggest that the populations driving DNRA and denitrification differ, with potentially little overlap.

Implications for DNRA and denitrification in on-site paddy soil
The results obtained herein and previous field observations in the same location suggest that DNRA is a substantial N cycling process in on-site paddy soil under practical management conditions. A previous field observation study --reported that soil NO 3 and NO 2 were continuously undetectable under anoxic field conditions (Itoh et al., 2013), suggesting that NO 3 -and NO 2 -produced by nitrification in the oxic layer are rapidly consumed by some pathway. Another field observation study that performed a metatranscriptome analysis detected nrfA transcripts in greater abundance in the reduced layer than in the surface, and many of these sequences were similar to those of Anaeromyxobacter and Geobacter (Masuda et al., 2017), suggesting that paddy soil in the reduced layer has the potential for DNRA largely by Anaeromyxobacter and Geobacter. By conducting controlled laboratory incubation experiments with 15 N tracer and gene transcript-clone library analyses, we herein demonstrated that DNRA occurred in rice paddy soils at a similar rate to denitrification in NO 3 -limiting (relative to organic carbon) anaerobic paddy soils and that Anaeromyxobacter-like and Geobacter-like nrfA expression enhanced DNRA activity. Based on these results, the paddy field may form more closed N-cycling than the nitrification-denitrification centric understanding; that is, NO 3 -produced by nitrification at the surface oxidized layer is reduced not only to gaseous N 2 via denitrification, but also to NH 4 + via DNRA, below the oxidized layer. Difficulties are associated with evaluating the extent to which DNRA and denitrification occur in on-site paddy soil because both are difficult to measure without the addition of a substrate ( 15 NO 3 -). The contribution of DNRA to the -depletion of NO 3 may be larger in an anaerobic environment with a large influx of readily degradable organic mat--ter, but with a small influx of NO 3 , leading to high org--C/NO 3 . An anaerobic environment around the rhizosphere, in which a large amount of root secretions is supplied (Kuzyakov and Blagodatskaya, 2015), may be one such environment. The soil redox potential may also affect the partitioning of NO 3 -reduction between DNRA and denitrification. However, a consensus has not yet been reached on the soil redox conditions under which DNRA and denitrification occur. While some studies demonstrated that DNRA occurs in more reduced soils/sediments than denitrification Patrick, 1978, 1981), others showed that DNRA was relatively insensitive to redox conditions (Pett-Ridge et al., 2006) and even occurred in aerobic-like environments (Morley and Baggs, 2010;Yang et al., 2017). Further studies are needed to identify the hotspots and hot moments of DNRA in on-site paddy soils (Kuzyakov and Blagodatskaya, 2015).
Rice paddy soil appears to have great potential for DNRA. Although research on DNRA in rice fields remains limited, DNRA has accounted for a significant proportion of NO 3 -reduction in the paddy soils tested to date (Pandey et al., 2018;, with some exceptions (Shan et al., 2016). The DNRA potentials of paddy soils (0.8-9.8 μg N [g soil] -1 d -1 from our second experiment, 0.42-3.20 μg N [g soil] -1 d -1 [Pandey et al., 2018;, 0.2-8.9 μg N [g soil] -1 d -1 [Buresh and Patrick, 1978], and 0.01-0.19 μg N [g soil] -1 d -1 [Shan et al., 2016]) are often larger than those of croplands, grasslands, and forests (<1.0 μg N [g soil] -1 d -1 [Rütting et al., 2011;Putz et al., 2018], except for 2.89 μg N [g soil] -1 d -1 in a humid tropical forest [Pett-Ridge et al., 2006]), although these potentials may largely depend on the experimental settings, such as the concentrations of substrates added. Moreover, Anaeromyxobacter and Geobacter, the main candidates responsible for DNRA, are widely predominant in Japanese paddy soils (Masuda et al., 2017). In addition, laboratory measurements of the potential for denitrification may overestimate denitrification and underesti--mate DNRA because only NO 3 is added as a substrate in many cases (Tiedje et al., 1989), which may enhance denitrification over DNRA. Even with the addition of organic carbon, the addition of NO 3 -relieves the NO 3 -limiting condition, which is also applicable to the present study. Recent studies suggested that high N fertilization suppresses DNRA, but enhances denitrification (Peng et al., 2016;Pandey et al., 2018;Putz et al., 2018), possibly due to an increase in NO 3 availability relative to organic carbon in soils. Rice paddy fields may exhibit the more efficient use of N fertilizers than upland crop fields (Okumura, 2002;Zhou et al., 2011), and strong DNRA activity in paddy soils may be the reason for this efficacy. The efficient enhancement of DNRA provides an opportunity to achieve the lowfertilization management of rice paddy fields.